What if automakers could design their next popular vehicle in minutes for a fraction of the cost? That’s what General Motors and the MIT Sloan School of Management explored in a recent study, using a neural network to analyze data collected from GM clinics to determine what makes an attractive vehicle. By inputting images and scores given by attendees, the AI software trained itself to output an aesthetic rating for a given image. The program was able to predict with remarkable accuracy how consumers would rate vehicles, achieving a 74% agreement rate.
Next, the program used this information to design new vehicles that would score high aesthetically, based on parameters such as “make it Chevy-like.” Interestingly, some of the designs generated by the AI were strikingly similar to vehicles sold by automakers in 2020, despite the fact that they were not in the automaker’s lineup in 2014.
While the use of generative technology in design is fast emerging, ArtCenter College of Design professor Chris Chapman cautions that it can lead to derivative designs that lack originality. He suggests that AI can be a useful tool in the design process, providing inspiration and unexpected ideas to work with, but it is ultimately up to the artists to inject originality into the final product.
Despite the software’s potential, it is not yet ready for prime time. The tool requires an expert to operate, and further development work will be needed to commercialize it. In the short run, however, MIT Sloan marketing professor John Hauser believes that the rating software can be helpful in guiding a design’s direction. By using AI to ensure that a design is aesthetically pleasing, designers can create iconic models that stand the test of time.
By Impact Lab